Research On Method Of Extracting Resident Trip OD Matrix Based On Mobile Phone Positioning Data

Urban traffic is the basis and premise of urban development, and is a bond that links economic activities in modern society together. If a city wants to develop rapidly, the most important thing is to solve the traffic problem. Solving the traffic problem mainly includes the diagnosis of the traffic problem, the planning for transport system and decision-making of traffic management policy, and one of its reference basises is resident trip OD survey (Origin Destination Survey), which means to investigate the overall situation of resident trip from the origin to the destination to learn the composition, traffic flow and direction of resident trip. Traditional method of resident trip OD survey mainly includes artificial inquiry and survey form, such as roadside survey, table survey, telephone survey and so on. These traditional methods need massive investment of manpower and financial resource, and it suffers from low sampling rate and poor accuracy. Methods of OD survey based on modern technology mainly include microvave detection, coil detection, video detection and floating car equiped with GPS. These advanced methods also have a few drawbacks, such as need to change the existing equipment, high cost of installation and maintenance, limited coverage rate, and so on. With the development of mobile communication technology, mobile phone penetration gradually increased. A large number of mobile phone users produced a mass of mobile positioning data in daily travel. The collectiong of mobile positioning data doesn’t need to change the existing mobile equipment, and has the characteristics of all-weather acquisition, high sampling rate and huge volume.Therefore, based on advantages of mobile phone positioning data, this paper study acquisition method of resident trip OD matrix based on mobile phone positioning data.The background, significance and present development in this filed at home and abroad of this research is shown firstly in this paper, then outline of this paper and relevant theoretical basis are introduced. Based on the analysis of mobile positioning data, data preprocessing is applied to deal with existing problems of positioning data, including removing defect data and drift data from positioning data and subdividing the positioning data into equal time intervals. Then a bi-level clustering algorithm for identifying stay point was proposed. In spatial level, clustering positioning data using DBSCAN algorithm and optimizing the clustering results based on k-nearest neighbor to get the candidate stay points. In temporal level, dividing the candidate stay point based on temporal property to obtain eventual stay point and trip chain of mobile users. Based on the similarity of clustering and traffic zone’s division, K-Means algorithm is used to cluster stay points, and the parameter k in K-Means algorithm is determined on the basis of the principles of traffic zone’s division. Then the trip chain is mapped to traffic zone to identify trip between different traffic zone, and resident trip OD matrix is obtained by using mathematical statistical methods. Finally, based on mobile phone positioning data of telecom users in Weifang, experiments are designed to verify the methods proposed in this paper, and we analyze the geographic distribution characteristics of OD traffic flow.